Strategic Choices in DSL Deployment

Grappling with DSL decisions? A tenured expert offers advice about intelligently selecting an operating standard that will make the most out of building broadband profits.

by Rick Gilbert [October 17, 2000]

Rick Gilbert is President and Chief Executive Officer of Copper Mountain Networks, Inc., a worldwide market-share leader in business-class DSL equipment. Gilbert hails from Portland, Oregon where he grew ADC Kentrox, a network access and multi-service WAN equipment provider from a $25 million business to a $144 million a year profitable division of ADC Telecommunications. Gilbert has also held senior management positions at Make Systems, Vitalink Communications Corp., and IBM Corp. before he arrived at Copper Mountain to take the company public in May 1999.

One of the first decisions providers face when building out networks that support Digital Subscriber Line is a fundamental choice between implementing access over Asynchronous Transfer Mode or Internet Protocol broadband solutions.

While this choice may initially seem purely a product decision based on the specific merits of one device over another, it is actually a strategic choice that will affect overall network operations, scalability, and service delivery well into the future.

Network and business planners should carefully consider both technical and strategic issues when deciding between these two solutions.

Not just a pretty voice box
Service providers today are building out networks with a view to delivering value-added voice, data, and content services, rather than merely providing Internet access.

To support multiple services such as voice, data, and videoconferencing via DSL, providers must implement intelligent systems with the quality of service (QoS) features required to prioritize and differentiate traffic by subscriber and application, which can be managed and expanded as quickly and simply as possible.

The choice between ATM and IP when it comes to DSL delivery should be evaluated with these criteria in mind.

Does a pricey QoS solution solve anything?
ATM evolved in voice-centric networks, and it has inherent QoS features that are implemented by establishing end-to-end virtual circuits (VCs) between the provider and the customer for each service delivered. To set up DSL via an ATM-based DSL access multiplexer (DSLAM), for example, a provider establishes two VCs: one for a voice line and one for a data line. The provider can then control the bandwidth and prioritize traffic individually for each VC.

But virtual circuits don't scale well. Suppose a DSL service provider uses ATM to deliver Internet access, voice, and virtual private networking (VPN) service to 100 subscribers. This provider must establish 300 separate VCs to provision service through an ATM DSLAM, and a human operator has to access the management system of each network component and manually set up and test each virtual circuit for every subscriber.

Multiply the provisioning issues across scores and potentially hundreds of service providers and millions of subscribers, and you have an entire network at the breaking point.

Analysts project that there will be more than 2 million DSL subscribers in North America by the end of 2000, and more than 7.6 million by the end of 2002. This exponential increase in subscribers, combined with a need to support multiple services per subscriber, would require millions upon millions of VCs. Neither provisioning personnel nor existing ATM equipment can support this impending explosion of circuits. ATM proponents tout switched virtual circuits (SVCs) as an answer to the sheer volume of work involved in provisioning individual VCs. But SVCs, while reducing the number of circuits, put a tremendous strain on signaling and route-processing resources. Reliance on SVCs also imposes large expenses on providers and users, who must change current CPE, LAN, and WAN infrastructures in order for the provider to integrate IP layer services with the ATM layer.

An ATM solution can resolve the QoS problem for multi-service delivery via DSL, but it is still excessively expensive to implement, scale, and manage.

To be IP, or not to be
The other solution to the DSL QoS problem is to use intelligent, IP-based DSLAMs, which leverage the inherent scalability and operational economy of an all-IP network.

Intelligent DSLAMs can use IP header information to implement Class of Service (CoS) capabilities, giving providers the same control over individual services that they get with ATM.

IP CoS mechanisms read information from packet headers and use it to prioritize traffic. The mechanisms assign each packet to a specified Forwarding Equivalence Class. Each service class specifies bandwidth, delay, priority, and other characteristics, and the intelligent IP DSLAM aggregates the packet flows and forwards them according to the service class of the packet.

The IP DSLAM's ability to aggregate and forward traffic according to service class eliminates the need for providers to set up and manage separate VCs for each service for each subscriber.

In addition, providers who use IP for DSL do not have to integrate a different protocol into an otherwise all-IP network. Service providers have built IP-based networks for flexible and efficient data delivery. Voice over DSL (VoDSL) solutions allow providers to offer voice services over these networks as well. By combining voice and data services in an intelligent IP DSLAM, providers can maintain a homogenous network that is easier to scale, less expensive to operate, and more reliable overall.

Concentrate on convergence
The decision about how to implement DSL service is also a strategic one that affects a provider's overall business model, and again, IP makes the most sense.

Service providers must quickly adopt best-of-breed solutions throughout their networks to quickly offer the new, converged services they need for market differentiation and profits. This means buying products that are supported by a large developer community and that can quickly be integrated into an existing infrastructure.

Because of its roots in core telecommunications networks, ATM is used by a relatively small number of carriers, and ATM products are developed by only a handful of companies. In contrast, there are dozens of companies in the IP networking industry all vying to deliver advanced solutions.

Based on the overwhelming support for IP networking solutions, service providers will find it far easier to buy and integrate best-of-breed solutions for IP than for ATM.

When evaluating the difference between ATM and IP DSLAMS today, service providers should ask whether it's worth introducing a second protocol — with all the conversion overhead that it entails and is supported by a handful of vendors — or whether the best move is to use the same protocol as the rest of the network, one which is supported by hundreds of vendors.

The choice should be clear.

—End